1 / 26

First Level Trigger using Associative Memories for CMS at Super-LHC

First Level Trigger using Associative Memories for CMS at Super-LHC. Fabrizio Palla INFN – Pisa Acknowledgments to A. Annovi, M. Dell’Orso, P. Giannetti, G. Parrini, M. Vos, F. Vasey Published on JINST 2 P02002. Some issues for SLHC Triggers @ 10 35.

mac
Download Presentation

First Level Trigger using Associative Memories for CMS at Super-LHC

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. First Level Trigger using Associative Memories for CMS at Super-LHC Fabrizio Palla INFN – Pisa Acknowledgments to A. Annovi, M. Dell’Orso, P. Giannetti, G. Parrini, M. Vos, F. Vasey Published on JINST 2 P02002

  2. Some issues for SLHC Triggers @ 1035 • ~400 Minimum Bias events/bx (50 ns) [16xLHC] • Occupancy • Degraded performance of algorithms • Electrons: reduced rejection at fixed efficiency from isolation • Muons: increased background rates from accidental coincidences • Larger event size to be read out • New Tracker: higher channel count & occupancy  large factor, but can be accommodated if digital readout chosen • Reduces the max level-1 rate for fixed bandwidth readout. • Implies raising ET thresholds on electrons, photons, muons, jets and use of less inclusive triggers • Need to compensate for larger interaction rate & degradation in algorithm performance due to occupancy Tracker Trigger for SLHC

  3. Move some HLT algorithms into L-1 or design new algorithms reflecting tracking trigger capabilities CMS Muon Rate at L =1034 cm-2 s-1 From CMS-DAQ TDR Note limited rejection power (slope) without tracker information Tracker Trigger for SLHC

  4. Tracking Trigger driving ideas • Design considerations: • Main usage for pT reconstruction • We need low occupancy and large lever-arm, rather than brilliant space-point resolution • If the Tracker material budget will not decrease by a sizeable amount, multiple scattering will drive the momentum resolution below ~10 GeV • Use binary readout (0/1) for each channel • Avoid duplication of signals to avoid fibers increase • Use readout signals for Trigger  move the whole data from the detector out in the barracks • Use established detector and electronics technologies to get a robust project, however • Push current pixel systems (see R. Horisberger, T. Rohe et al in SLHC CMS workshops) • Expand existing (and working) Tracking Triggers (CDF SVT) • Profit of miniaturization in electronics • More radiation hard optical fibers technologies (F. Vasey…) Tracker Trigger for SLHC

  5. Mitigate from CDF SVX approach Main problem: input Bandwidth (see later) divide the detector in thin fsectors. Each AM searches in a small Df OFF DETECTOR 1 AM for each enough-small Df Patterns Hits: position+time stamp All patterns inside a single chip N chips for N overlapping events identified by the time stamp Data links Event1 AMchip1 Event2 AMchip2 Event3 AMchip3 EventN AMchipN Tracker Trigger for SLHC

  6. The pattern bank is flexible • set of pre-calculated patterns: • can account for misalignment • changing detector conditions • beam movement • … The Pattern Bank ... Pattern matching in CDF (M. Dell’Orso, L.Ristori 1985 -..) The Event Tracker Trigger for SLHC

  7. On the opposite side: FPGA for the same AMchip P. Giannetti et al. “A Programmable Associative Memory for TrackFinding”, Nucl. Intsr. and Meth., vol. A413/2-3, pp.367-373,(1998). In the middle: Standard Cell 0.18 mm (INFN-Pisa) 5000 pattern/chip AMchip L. Sartori, A. Annovi et al., “A VLSI Processor for Fast Track Finding Based on Content Addressable Memories”, IEEE Transactions on Nuclear Science, Volume 53, Issue 4, Part 2, Aug. 2006 Page(s):2428 - 2433 NEXT: NEW VERSION For both L1 & L2 AM chips from 1992 to 2005 • (90’s) Full custom VLSI chip - 0.7mm (INFN-Pisa) • 128 patterns, 6x12bit words each • 32k roads / wedge • F. Morsani et al., “The AMchip: a Full-custom MOS VLSI • Associative memory for Pattern Recognition”, IEEE Trans. on Nucl.Sci., vol. 39, pp. 795-797, (1992). Tracker Trigger for SLHC

  8. TOB TEC TIB TID PD CMS Tracker Trigger Layouta feasibility study • As a study, limited to the same barrel region as for the current inner Silicon Strip Tracker barrel (TIB) • TOB should also be included, for much reduced number of hits, however not in this study • 4 layers with radii at 26, 34, 42, 50 cm from the beam line. • Barrel coverage up to |h|<1.5 110 cm Tracker Trigger for SLHC

  9. A possible Inner Tracker Layout • Granularity and costs driven approach • Long fibers (~100 m) and high speed (>2.5 Gbps) to extract signal off detector • Pixel System for radius at 26 cm • Single sided pixel detectors, n+ on p – Silicon (Czochralski) Pixel area ~ 160 mm x 650 mm • Sensor area 2 (r-f) x 8 (z) cm2 • 1 - 2 fibers/module for 5 Gbps • Pixel System for radii at 34, 42, 50 cm • Silicon strips (actual) have sensor element area of 10 to 15 mm2 • 10 fold increase in the luminosity would need a 10 fold decrease of it • Large elongated pixels of 200 mm x 5 mm • Sensor area 6 (r-f) x 12 (z) cm2 • 3 - 4 fibers/module for 5 Gbps Tracker Trigger for SLHC

  10. Sensors and signal routing Large radii sensors 6 cm Fiber(s) hybrid Readout chips 200x5000 mm2 5 cm Current CMS pixel sensors 8 cm 2 cm fibers Fiber(s) 160 x 650 mm2 32 bits sufficient to locate pixel position in the sector and bx time stamp Small radius sensors From an idea of R. Horisberger http://agenda.cern.ch/fullAgenda.php?ida=a036368 Tracker Trigger for SLHC

  11. Subdivide the detector in many f sectors Keep data volume limited in each sector Combine information from at least 3 layers out of 4 in each sector Momentum resolution of ~ few (<10)% at 10 GeV/c Granularity driven by the minimum measurable pT for triggering purposes, without loosing efficiency ~80 f sectors at the innermost radius Df~ 4.5° matches to a module of 2 cm width Well covering the bending of a track of 5 GeV pT and above Larger f sectors with increasing radii Match the sensors widths Sector segmentation 50 cm 42 cm 34 cm 26 cm Tracker Trigger for SLHC

  12. Occupancy studies • GEANT4 simulation of pixelized tracking layers • Simulated 3500 minimum bias using latest Pythia settings events and group into chunks of 100 events per bunch crossing and 250 t-tbar events • Use current CMS layout (and material budget) but different sensors granularity Current links in CMS TIB Silicon Strip: 2000 @ 26 cm - 2600 @ 34 cm ª average number on minimum bias events, t-t will contribute on average<<1 hit/det – 12.5 ns bx It will only double for 25 ns *32 bits/hit †for a data link speed of 5 Gbps – does not change with bunch spacing Tracker Trigger for SLHC

  13. Number of hits/module/bx R=34 cm R=26 cm R=42 cm Could accommodate fluctuations using buffers on the detector, or increasing the number of fibers Tracker Trigger for SLHC

  14. Layer 0: ~25 fibers bringing ~40 Hits/12 ns Parallel IN Serial OUT AM EV0 1 Hit/10 ns From other layers Parallel IN Serial OUT AM EV1 1 Hit/10 ns From other layers Distribute hits into different sets of storage units depending on EVent # . . . . . . Parallel IN Serial OUT AM EV40 1 Hit/10 ns From other layers 1 FPGA Conceptual design From Detector Tracker Trigger for SLHC

  15. 6 cm hybrid Readout chips 50 mm x 10 cm 10 cm Pushing to outer radii • Extrapolating the previous layout to outer radii could improve the momentum resolution increasing the lever arm. • Z coverage up to ±130 cm (for TIB-like configuration up to ±55 cm) • In order to cover the bending of muons of at least 10 GeV pT need to consider trigger sectors in phi made by two sensors. • Proposed cell dimension ~6-8 times smaller wrt the current (CMS) one • See also next slide for the pitch choice Current links in CMS TOB Silicon Strip: 1300 @ 60 cm – 6000 overall Tracker Trigger for SLHC

  16. Using the cluster width to reduce trigger data rate G. Parrini INFN - Florence MIP R-phi plane track width • In this region, using 50µm pitch, about 3% of the total particles leave cluster sizes with ≤2 strips: ~ 3 GeV pT • It could be used for reducing the information for triggering purposes of more than one order of magnitude, leaving the rest (AM) untouched. • It would need dedicated ASIC on silicon module. • Once reduced to ~100 KHz, it would only need few fast readout links to readout the entire Tracker ∆R R Tracker Trigger for SLHC

  17. Efficiency, latency and power • To increase efficiency in f coverage could send the same fiber to more than one sector • Could accommodate more external layers if needed, to reduce ghosts and increase momentum resolution • Latencies • ~100 m fiber 300 ns [24 bx] • Switch + AM ~1ms [80 bx] • Sensor read-out latency budget should be less than ~80 bx • TOTAL ~ <200 bx [1.75 ms] • Power consumption (R=26 cm, largest density) from laser drivers • Largest consumption innermost modules • ~1W/module  ~100 mW/cm2. • Current systems (CMS) from pixel ROCs ~ 200 mW/cm2. Tracker Trigger for SLHC

  18. Board dimensions and costs (P. Giannetti et al) Board dimensions • About 30,000 patterns per AM chip required • Needed ~ 80 boards with ~40 AM chips each • 3200 AM chips • The current AM for CDF holds ~5000 patterns/6 planes in 0.18mm technology • If developed in the 90 nm technology one could accommodate ~4 times more patterns/AM chip hence 30,000 for 4 planes • A batch with 3000 AM chips (yield of 75%) costs 30k$ Tracker Trigger for SLHC

  19. Muon Trigger Rate • Estimate of the L1 Trigger rate for different pT • Assume very simple Tracker Trigger finding algorithm • No isolation required • Correlate with estimated L1-Muon alone Preliminary Tracker Trigger for SLHC

  20. Conclusions – I • A conservative approach to a Tracker based Trigger for SLHC based on the existing technologies has been presented • Very precise “pixel” layers at intermediate radii (25 to 50 cm) • Relies upon “standard” detector technologies • Limits the data volume to be transferred • Readout buffers could be useful to accommodate fluctuations in the number of hits/event • Small power density compared to innermost layer based approaches • Very simple digital readout • Simulation suggest usage of the outermost layers (TOB) could be used • Higher lever arm • Lower data rate especially when cutting low pt tracks using cluster width information • High speed radiation-hard links very likely available for SLHC • Allow off-detector track reconstruction • Use same fibers for Trigger as well as for read-out • Data link reduction • Challenging, need further R&D Tracker Trigger for SLHC

  21. Conclusions – II • Full track reconstruction using parallel processing • Use • current efforts to extend positive CDF (SVT) experience • Foreseen electronic miniaturization • Heavily based on fast switches. Need R&D • Used in conjunction with other “traditional” (e,m, calo) L1 primitives allows to reduce L1 rate to less than 100 KHz • Could put additional processing to look for di-lepton resonances at L1 and more exotica • Production costs not driven by the Trigger boards • First starting point, but … • Need further R&D • Need to be integrated with detector and with mechanics • Could we reduce the number of links? • Complements inner layers Trigger approach Tracker Trigger for SLHC

  22. spare slides Tracker Trigger for SLHC

  23. Trigger design (start from Physics!) • Main problem is the large number of tracks/bx • dN/dy|y=0 ~ 6 • @S-LHC ~2000 tracks/bx in |h|<1.5 • But only a few ‰ have pT>5 GeV/c and less if pT>10 GeV/c • In principle, triggering on these tracks would give a reduction from 80 MHz to 80 KHz or less, if the fake rate is under control Tracker Trigger for SLHC

  24. Layer 0: ~25 fibers bringing ~20 Hits/12 ns 1 Hit/10 ns AM EV0 . . . . . . From other layers 1 Hit/10 ns AM EV1 . . . From other layers Distribute hits into different sets of FIFOs depending on EVent # . . . 1 Hit/10 ns AM EV20 . . . From other layers Tracker Trigger for SLHC 1 FIFO/fiber

  25. . . . Layer 0: ~25 fibers bringing ~20 Hits/12 ns 1 Hit/10 ns AM EV0 . . . From other layers 1 FIFO/ fiber 1 Hit/10 ns AM EV1 From other layers Distribute hits into different sets of registers depending on EVent # . . . FPGA 1 Hit/10 ns AM EV20 From other layers Tracker Trigger for SLHC

  26. Layer 0: ~25 fibers bringing ~20 Hits/12 ns Parallel IN Serial OUT AM EV0 1 Hit/10 ns From other layers Parallel IN Serial OUT AM EV1 1 Hit/10 ns From other layers Distribute hits into different sets of storage units depending on EVent # . . . . . . Parallel IN Serial OUT AM EV20 1 Hit/10 ns From other layers 1 FPGA Tracker Trigger for SLHC

More Related